You rarely think about network performance until something doesn’t work as expected. The complaint may be a slow file transfer or an interrupted videoconference. In any case, you want the problem fixed immediately and you don’t want it to happen again. Georgia Tech is working on ways to find and fix network performance glitches quickly.

The process for resolving a computer network issue has typically relied on an informal people network. The end user calls a support person who may have to call another support person to troubleshoot the issue and then track down the appropriate technician to work on the network. The process is hit or miss.

Now, researchers and students at the Georgia Tech Research Network Operations Center (RNOC) are developing a system that will identify network performance issues quickly so that they can be resolved quickly, preferably before the end user even notices them.

In spring 2005, the RNOC team, led by Dr. Russell Clark, College of Computing, and Dr. Warren Matthews, OIT, installed a campuswide network performance monitoring and recovery system, called CPR, at Georgia Tech. The CPR system is a collection of devices and programs dedicated to assessing network performance.

More than 60 computer nodes are monitoring network connections and services at buildings on the main campus and at Georgia Tech satellite campuses. It’s important to note that the monitoring devices -- the computer nodes -- are distributed as close to the end user as possible. The performance data represents the user’s experience in the field rather than the network’s performance at a centralized location. Dr. Clark says there is no commercial product or service that compares to the CPR system.

The CPR data sample below shows the network performance between the MARC building and the central servers in the Rich building. It indicates the experience a student or faculty member would receive collecting their email or checking the Georgia Tech Web site.

The data on this graph indicates very good network performance

The RNOC team is already using the CPR system to troubleshoot Georgia Tech network issues. Next, they want to make the data available to more people on campus. One group of students is already developing a Web site to display CPR data. In the fall, another student group will work on automating the generation of CPR reports. The challenge is to automatically process the data and summarize it in performance reports that are useful to people who are not network engineers. “We want to empower researchers in the field to diagnose the network performance problems in their buildings,” said Dr. Clark.

Dr. Matthews is leading an initiative, called the Georgia Measurement and Monitoring system (GAMMON), to extend the CPR model to research centers and universities across Georgia, starting with the University of Georgia, Emory, and Georgia State University. The GAMMON project promotes collaboration as a way for network engineers to help researchers resolve common issues, such as how to speed up the transfer of large data sets.

Georgia Tech is also extending the CPR model to the campus wireless network, using monitoring programs on wireless access points and computer nodes, and an application in development that end users will install on their wireless devices. The “WeedEater” program will show how well the wireless network is performing as measured on the wireless user’s device.

The CPR system at Georgia Tech was relatively inexpensive to set up. The computer nodes are surplus PCs configured with a Linux operating system. Installing the system was a real-world learning experience for Dr. Clark’s network engineering students.

For the latest CPR documentation, including a Status Report and Roadmap, go to http://www.rnoc.gatech.edu/cpr/documentation.html.

More information

For more information, contact Russ Clark (russell.clark@cc.gatech.edu) or Warren Matthews (warren.matthews@oit.gatech.edu). View the RNOC Web site at http://rnoc.gatech.edu/.

Top